`Artificial Intelligence' Solves a Mystery

ARCHAEOLOGY, artificial intelligence, and cryptography have little to do with one another, and is that at once both the charm and the problem with Alexander Tzonis's novel ``Hermes and the Golden Thinking Machine.'' Hermes Steganos is a young professor of archaeology at Harvard University, by all accounts a genius who has succeeded in applying the science of artificial intelligence (AI) and his knack for breaking codes to the decryption of ancient texts.

The Golden Thinking Machine is a priceless artifact, stolen by Hermes's uncle - also an archaeologist - from a recent dig. Hermes sees the machine briefly at the start of his year-long sabbatical in Athens. Hours later Hermes's uncle is murdered and the Golden Thinking Machine stolen. Hermes is the prime suspect. In order to prove his innocence, he must find the killer.

The plot quickly unfolds in this charmingly academic novel about the nature of thought and the history of machine intelligence. The uncle's circle of friends includes archaeologists, spies, and mystics, all of them erudite, nearly all of them with a secret desire to study - what else? - artificial intelligence.

As Hermes seeks them out, for information or interrogation or help, the subject of the discussion almost invariably turns to search strategies, logical decision trees, problem-solving programs, and reasoning by analogy. As his contacts become more exotic, Hermes becomes more confused.

Meanwhile, back in his apartment, Hermes's beautiful and equally brilliant cousin Nina, herself a student at a nearby university, has a plan.

Using the experimental AI laptop computer that Hermes has brought with him from the United States, Nina is writing an AI ``expert system'' that will solve the crime. Although she has no formal knowledge of the subject, she is a fast learner and has one of the world's best teachers as her instructor. Nina spends nights talking with Hermes and her days hacking.

Nina's involvement lets Tzonis recapitulate the last 30 years of AI research. Each part of the mystery focuses on some significant tenet of AI, starting with the representation of knowledge and ending with reasoning by analogy. Nina shares the revelations, glories, and failures of such AI greats as Marvin Minsky and Patrick Winston, experiencing the promise of AI and finally realizing why the science of AI has still not yet come up with the goods.

At times, ``Hermes and the Golden Thinking Machine'' has the feel of a professor's nightmare gone haywire - not surprising, since Tzonis taught at Harvard from 1967 to 1981. Thankfully, he never uses his characters' interest in AI as an excuse to lecture the reader about the field. Instead, the novel uses artificial intelligence the way that ``Mutiny on the Bounty'' uses sailing: The subject is an ever-present backdrop, but many of the references remain unelaborated.

For the reader knowledgeable about AI, the novel comes across as a coffeehouse discussion among greats in the AI field, each making use of the ideas and methods for their own ends.

For the reader unfamiliar with the discipline, the mystery makes a wonderful and painless first introduction to artificial intelligence. A reading list in the back shows inquisitive readers where to go for more rigorous treatments of the subject.